Efficient Space Skipping and Adaptive Sampling of Unstructured Volumes Using Hardware Accelerated Ray Tracing

Abstract

a) (b) (c) (d) Figure 1: Performance improvement of our method on the 278 million tetrahedra Japan Earthquake data set. (a) A reference volume ray marcher without our method, at 0.9 FPS (1024 2 pixels) on an NVIDIA RTX 8000 GPU. (b) A heat map of relative cost per-pixel in (a). (c) and (d), the same, but now with our space skipping and adaptive sampling method, running at 7 FPS (7× faster). ABSTRACTSample based ray marching is an effective method for direct volume rendering of unstructured meshes. However, sampling such meshes remains expensive, and strategies to reduce the number of samples taken have received relatively little attention. In this paper, we introduce a method for rendering unstructured meshes using a combination of a coarse spatial acceleration structure and hardware-accelerated ray tracing. Our approach enables efficient empty space skipping and adaptive sampling of unstructured meshes, and outperforms a reference ray marcher by up to 7×.

Point Containment Queries on Ray-Tracing Cores for AMR Flow Visualization

“…Reformulating one's algorithm is, of course, not always possible or efficient. For example, the problem domain must be embeddable in R 3 , and in general, the problem must be mappable to a ray tracing problem without introducing significant overhead. In the previous example of the radius point queries [1], the authors observed that a fixed radius point query over a set of particles could be reformulated as a ray-tracing problem by using Ray-traced point containment queries.…”

“…Point containment queries with RT cores have been proposed by several researchers, including the aforementioned force-directed graph drawing algorithm by Zellmann et al [1]. Even more closely related to our approach is the work by Morrical et al [3] who used point containment queries to evaluate the particle density at arbitrary sampling positions x ∈ R 3 given an underlying unstructured grid representation with finite ele-ments such as tetrahedra or hexahedra.…”

A non-exponential transmittance model for volumetric scene representations

“…In its hierarchical form, it allows to efficiently bound the per-pixel rendering complexity by choosing the prefiltered scale proportionally to the pixel's footprint. The rendering of a volumetric representations can be accelerated by using sparse data structures [Crassin et al 2009] and empty space skipping [Hadwiger et al 2018;Morrical et al 2019]. Beyer et al [2015] provide an overview of efficient GPU volume rendering techniques used in scientific visualization.…”

Differentiable Direct Volume Rendering

“…In prior works, strategies for optimal sampling in screen-space have been proposed, for instance, based on perceptual models [4], image saliency maps [39], entropy-based measures [56], temporal history [29], or using neural networks [53,54]. Other approaches adaptively change the sampling stepsize along the view rays to reduce the number of samples in regions that do not contribute much to the image [6,9,23,32]. DiffDVR's capability to compute gradients with respect to the stepsize gives rise to a gradientbased adaptation using image-based loss functions instead of gradientfree optimizations or heuristics.…”